Research map / 3D vision / Gaussian splatting
Long-LRM: Long-sequence Large Reconstruction Model for Wide-coverage Gaussian Splats
Chen Ziwen, Hao Tan, Kai Zhang, Sai Bi, Fujun Luan, Yicong Hong, Li Fuxin, Zexiang Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace optimization-based 3D GS with feed-forward Long-LRM at 32 views | Long-LRM · DL3DV-140 (32 views) | reconstruction time: 13min -> 1sec | Improved |
| add 10 post-prediction optimization steps | Long-LRM · DL3DV-140 (32 views) | PSNR: 24.10 -> 25.60 | Improved |
| Long-LRM versus optimization-based methods on perceptual quality | Long-LRM · DL3DV-140 (32 views) | LPIPS: 0.205 -> 0.254 | Hurt |
| replace all-Transformer with hybrid Mamba2+Transformer blocks (ablation) | Long-LRM · 32 views 256 res | PSNR: 24.28 -> 26.82 | Improved |
| use hybrid instead of all-Transformer at 4 views (ablation) | Long-LRM · 4 views 256 res | PSNR: 21.13 -> 21.58 | Improved |
| add token merging with token dimensionality 256 to 1024 (ablation) | Long-LRM hybrid · 32 views 512 res | iteration time (sec): 11.5 -> 4.0 | Improved |
| add token merging with token dimensionality 256 to 1024 (ablation) | Long-LRM hybrid · 32 views 512 res | PSNR: 28.16 -> 27.46 | Hurt |
| add opacity regularization loss (ablation) | Long-LRM · 4 views 256 res | % Gaussians with opacity>0.001: 99.2 -> 68.3 | Improved |
| add depth supervision from DepthAnything (ablation) | Long-LRM · 4 views 256 res | PSNR: 20.96 -> 21.25 | Improved |
| add opacity and depth losses on two-view RealEstate10K (ablation) | Long-LRM · RealEstate10K (2 views) | PSNR: 28.54 -> 28.44 | No clear effect |
| concentrate transformer blocks (3T21M or 21M3T) instead of evenly distributing them | Long-LRM hybrid · 4 views 256 res | training stability | Hurt |
| fine-tune Long-LRM to 2D GS | Long-LRM · ScanNetv2 zero-shot | PSNR: 19.25 -> 23.27 | Improved |
Later papers that compared with this method
7 later accepted papers report Long-LRM as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Multi-view Pyramid Transformer: Look Coarser to See Broader (CVPR 2026 · better on 4 of 4 shared results)
- iLRM: An Iterative Large 3D Reconstruction Model (CVPR 2026 · better on 2 of 3 shared results)
- SparseSplat: Towards Applicable Feed-Forward 3D Gaussian Splatting with Pixel-Unaligned Prediction (CVPR 2026 · better on 1 of 1 shared result)
- SceneTok: A Compressed, Diffusable Token Space for 3D Scenes (CVPR 2026 · better on 0 of 1 shared result)
- ReSplat: Learning Recurrent Gaussian Splatting (ECCV 2026 · better on 1 of 1 shared result)
- MegaSynth: Scaling Up 3D Scene Reconstruction with Synthesized Data (CVPR 2025 · better on 3 of 3 shared results)
- Stable Virtual Camera: Generative View Synthesis with Diffusion Models (ICCV 2025 · better on 0 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Scaffold-GS 30k — Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024) · better on 3 of 4 shared results
- MVSplat — MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (ECCV 2024) · better on 1 of 1 shared result
- GS-LRM — GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LRM: Large Reconstruction Model for Single Image to 3D (ICLR 2024 · similarity 0.90)
- DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos (NeurIPS 2025 · similarity 0.89)
- VastGaussian: Vast 3D Gaussians for Large Scene Reconstruction (CVPR 2024 · similarity 0.88)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.88)
- 4D-LRM: Large Space-Time Reconstruction Model From and To Any View at Any Time (NeurIPS 2025 · similarity 0.88)
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
